Loading observation history…
SHIPWATCH
A closer look at who’s shipping.
Follow the websites you’re curious about. See their release history, find their rhythm, and share what you notice.
The index
| # | WEBSITE | COLLECTION | RELEASES | LAST SHIP | STATUS |
|---|---|---|---|---|---|
| Loading websites… | |||||
Anyone can add a public website. The YC directory is a starter set. How it works
Recently observed
Listening for the next ship.
New releases appear here after confirmation.
No
backfilled or estimated activity.
How it works
01
Look every five minutes
We fetch each eligible public page and compare its first-party JavaScript content, including linked chunks within a bounded budget. Checks are spread out, with backoff for failures.
02
Confirm the change
The first successful check establishes a baseline. A different content fingerprint must persist on a follow-up check before we record it. Deployment IDs, known Sentry debug metadata, and hashed asset references do not count by themselves. Multiple changes within the same five-minute interval count once per website. Previously seen fingerprints and rollbacks don't add to the count.
03
Keep the limits visible
These are observed frontend releases, not commits, backend deployments, or employee working hours. A/B tests and regional rollouts can still introduce noise. Multiple releases between checks can be missed.
04
Compare what we actually saw
Rankings use 7- or 30-day windows. New companies have shorter observation histories. Every profile shows its baseline date and successful checks. Timestamps use your device’s timezone. Rate calculations use completed UTC days.
YC directory data comes from the unofficial yc-oss API. Active and public companies with usable URLs are eligible. Marketing websites may release independently of the product. We inspect up to 32 first-party assets per page, including imports and Vite-listed styles, with a 20 MB response budget. Conditional requests reuse unchanged content. Uninspected routes and third-party CDN assets can be missed. Older deployment-based observations are excluded unless their historical content can be reverified.
Finding subdomains
We follow public links on the submitted website and related hosts. A model suggests which pages look like product apps; you choose what gets monitored. We do not enumerate every subdomain or discover private apps.
Why a website may not be checked successfully
Some sites have no observable first-party JavaScript, disallow checks in robots.txt, block requests, or return errors. Failed checks retry with backoff. A domain count means selected monitored hosts—not all subdomains a company owns.